I have a workflow that involves saving contours out as .ply files when running client-server on HPC systems.
Unfortunately, I’m frequently unable to do so because the poly count is above 2.1 billion and from what I understand, the MPI code is still using 32-bit messages. I can usually find a way around the problem when it comes up, but not always in a satisfactory manner.
I’ve been told been told by Andrew Bauer that you guys are aware of the issue. So I guess I’m just asking if you guys are planning to work on this in the near future, or if it’s a bit lower on the priority list for now?
Hi David. That’s a lot of polygons in a .ply file!
I’m not aware of plans to address that particular issue in the near future, I’m afraid.
Can you refresh my memory on this – did we think that this was this a limitation inside the parallel implementation of the PLY writer? If it was, the new implementation, possibly available in PV 5.7 but definitely available in PV 5.8, shouldn’t have this limitation.
Something made us think that it had to do with MPI message lengths. Unfortunately, the discussion was over a month ago and I can’t remember the exact error message right now.
Next time we’re working on a larger calculation, I’ll test out various version of Paraview and let you know.
I’ll post the message I get in here next time I see it, just for everybody’s reference.